This study is designed to systematically map the evolution of cooperation in a game where actors can avoid those they do not trust. It departs from a long tradition in game theory, in which studies of the Prisoners' Dilemma rested on the assumption that actors were not free to walk away from undesirable partners. Computer simulations are used to systematically explore the conditions under which actors learn when and with whom to interact, to be later tested in laboratory experiments with human subjects. The computer simulation uses artificial neural networks to model a self-organizing system of interdependent, co-adaptive actors. The results are expected to produce a learning theory model of adaptation to changing conditions, relevant to the individual participant, in contrast to current theories based on the Prisoners' Dilemma, which assume that adaptation occurs through changes at the population level. %%% The proposed research will make fundamental contributions to social theory by offering new insights into the ways individuals learn to handle social dilemmas . It makes comparably innovative methodological contributions through the use of artificial neural networks. These are also used to provide a more realistic construction of social dilemmas that promises to be applicable outside the laboratory.
